图像非广延高斯熵分割的阈值选取方法  

Threshold selection for image segmentation based on nonextensive Gaussian entropy

作　　者：聂方彦[1,2] 李建奇[1,2] 张平凤[2] 屠添翼[1,2] NIE Fang-yan;LI Jian-qi;ZHANG Ping-feng;TU Tian-yi(Hunan Province Cooperative Innovation Center for the Construction & Development of Dongting Lake Ecological Economic Zone,Hunan University of Arts and Science,Changde 415000,China;College of Computer and Electrical Engineering,Hunan University of Arts and Science,Changde 415000,China)

机构地区：[1]湖南文理学院洞庭湖生态经济区建设与发展湖南省协同创新中心,湖南常德415000 [2]湖南文理学院计算机与电气工程学院,湖南常德415000

出　　处：《激光与红外》2020年第12期1504-1511,共8页Laser & Infrared

基　　金：湖南省教育厅科学研究重点项目(No.18A368);湖南省教育厅科学研究优秀青年项目(No.19B393);湖南省高等学校省特色学科项目(湘教通〔2018〕469)资助。

摘　　要：图像分割是实现图像目标与背景分离的基础和关键,其中阈值化是最为流行和有效的技术之一。考虑到图像处理作业场景的复杂境况,实时有效的阈值分割方法的提出一直是研究和应用中的极大挑战。针对复杂的工业无损检测图像及红外图像目标分割问题,基于非广延高斯熵提出一种新的图像阈值分割方法。对于图像中像素灰度级不均衡、不规则的随机分布特点,提出方法体现了较好的处理能力。在与实践常用方法的比较实验中,实验结果表明了提出方法的较优异性能和推广应用前景。Image segmentation is the foundation and key to the separation of image target and background.Among these,thresholding is one of the most popular and effective techniques.Considering the complexity of the scenario in image processing,it is a great challenge to propose a threshold segmentation method with higher real-time and effectiveness.Aiming at the complex problem of segmentation for industrial nondestructive testing image and infrared image,a new image threshold segmentation method is proposed based on nonextensive Gaussian entropy.For the characteristic of uneven and irregular distribution of histogram,the proposed method shows a good ability to deal with these problems.In the comparison experiments with some methods,the results indicate that the presented method has better performance and broad application prospects.

关 键 词：图像分割 直方图阈值化 复杂图像 非广延高斯熵 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]